Guido Calabresi, Concerning Cause and the Law of Torts: An Essay for Harry Kalven, Jr., 43 U. Chi. L. Rev. 69 (1975)
. . . .

In recent years, tort law has been reexamined with certain ends or functions consciously in mind.  Commentators have asked how the law of torts determines what injuries are worth avoiding, how it controls what categories of people bear the burden of those injuries that do occur and the related burden of avoiding those injuries deemed worth avoiding, and how it serves to encourage or require the spreading of such burdens. Finally, commentators have considered how these goals relate to the concept and language of justice – the test, however vague and uncertain, by which any area of law must be judged….

[W]hat role does causal language play in the achievement of certain goals that have come to be accepted as crucial to the law of torts? To accomplish this function-oriented task it is useful to distinguish three concepts of "cause": "causal link," "but for cause," and ''proximate cause."

The first concept, as I shall use it, is entirely predictive and empirical. There is a causal link between an act or activity and an injury when we conclude on the basis of the available evidence that the recurrence of that act or activity will increase the chances that the injury will also occur. . . . . Mr. Chief Justice Burger's bicycle riding and drag-racing on Washington streets by teenagers are both causally linked to the accident the Chief Justice had. This is true regardless of the value society places on either riding bicycles or drag-racing on the public streets. . . .
 
Obviously there are an infinite number of acts or activities that  are causally linked to every injury. Some may be viewed as socially desirable, some as neutral, and some as undesirable or even despicable. Getting up in the morning and going out increases the chances of many injuries – some of which are bound to occur. So does driving or practicing medicine. . . .

  The second concept of cause is the familiar one of but for or  sine qua non  cause. A but for cause . . . is any one of many acts or activities without which a particular injury would not have occurred. . . .

The final concept I shall consider is that of proximate cause. This concept, again familiar enough, will be used in a largely conclusory  sense. It will embrace those presumably causally linked,  usually (but by no means always) but for causes to which, in the absence of certain specified defenses, a particular legal system wishes to assign at least partial responsibility for an accident. . . .

I.    THE REQUIREMENT OF CAUSATION IN TORTS: CAUSE AS A FUNCTIONAL CONCEPT

A.    Compensation Goals

Tort compensation goals may be described in rough terms as follows: since people not only value a lump sum of money differently, but place a different value on each subsequent dollar as well,  the total impact of an injury may be diminished by an appropriate allocation of its burden. If, in general, a large lump sum burden is more onerous when borne by one person than it would be if divided  among many, then one function of tort law may be to ensure that, consistent with other goals, injury burdens are spread. Moreover, if even heavy burdens are less onerous when borne by certain wealth categories rather than others, then another function of tort law, again consistent with other goals, may be to allocate injury burdens to those wealth categories able to bear them with relative ease. . . .

1. Spreading.  If spreading of injury losses were the only goal of tort law, there would be no point at all in requiring, as a prerequisite to liability, a causal link between an act or activity and the  injury. . . .

The same is true of the requirement of a but for relationship between an act or activity and the injury. The fact that some behavior was or was not a but for cause of an injury says nothing about the ability of those who engaged in that behavior to lessen the impact of the injury by spreading its burdens. Thus, any requirement that responsibility be accompanied by a but for link, just like any requirement that it be accompanied by a causal link, must find its  justification in tort functions other than the spreading of injury  losses.

If spreading were the only goal of tort law, a social insurance fund, raised through taxes assessed on a per capita basis, would be the optimal mechanism for achieving it. Such a compensation system, designed to spread all accident losses to the maximum degree  possible, would not ask causal questions. . . .

The requirement of proximity, however, may have some role to play in regard to spreading aims. If . . . . if other tort goals, yet to be discussed, make the requirements of causal link or but for relationship desirable, then the additional  requirement of proximity might be used to pick relatively good "spreaders" from among those potentially liable actors . . . .

Some elements in proximate cause, like the requirement of foreseeability . . . are clearly germane to ability to spread  losses. . . . Since spreading ability is . . . crucially linked to insurance coverage, it is not hard to see how foresight, as an element in proximate cause, is itself linked to spreading ability.

Similarly, that aspect of proximate cause concerned with the  "size and nature of the damages" is closely linked to the ability of  the potentially liable party to "self-insure" . . . .  [If] findings of proximate cause turn or appear to turn on the size of the damages and the ability of the defendant to  bear them and still keep going[, then] . . . proximate  cause furthers the goal of spreading. . . .

Again, one should recognize that the concept of proximate cause may at most serve as a limit, a device for selecting the best  spreader out of a group of potentially liable actors who were originally identified on the basis of causation principles irrelevant to spreading. . . .   To repeat, a system of accident law designed with nothing but spreading in mind would be a social insurance system funded by  general taxes; such a system would necessarily do away with any  requirement of causation, including that of proximity. Thus, to find  a justification for requirements of causation in torts, it is necessary  to look beyond spreading goals.

  2. Wealth Distribution or "Deep Pocket." For much the same reasons that spreading functions do not help explain the requirement  of causation in torts, distributional goals also fail as  explanatory principles. . . .  The fact that certain  activities increase the danger of future accidents says nothing about the relative wealth of the parties involved. . . .

[I]t is possible to redefine causal linkage and but for relationships so that they run, not from "action" to "injury," but from wealth, status, caste or even generalized merit . . . to injury.  Causal  requirements could be shaped as follows: "in order to impose liability we must be able to say that, but for his wealth the injury would  not have occurred" or "we require that his wealth increase the chances of his engaging in activities that in turn increase the probability of future occurrences of this injury." . . . . [But] tort law has never been concerned with these types of  causal relationships. . . .

The role of the proximity requirement is more ambiguous. Here  too, since it serves only as a limitation, a device for selecting among  potentially responsible parties, it cannot be used to pick out that party who, in distributive terms, is best suited to bear the burden.  It may be used, however, to select a relatively good loss bearer in  terms of wealth distribution goals from among a group of parties who are deemed potentially liable in terms of other tort goals. . . . [I]t may well be that distributional  considerations are implicit in how juries, and perhaps even courts,  apply the test of proximity. . . .

B. Deterrence Goals

Deterrence goals in tort law may be described as those which  seek to minimize the sum of injury costs and safety costs. These  goals are to be achieved not by mitigating the burden of costs that  have already occurred, but by creating incentives so that people will avoid those future injuries worth avoiding and thus achieve an optimal  trade-off between safety and injury in a world where safety is not a free good, and hence injury is not a total bad. . . .

1. Collective or Specific Deterrence. One way of achieving the appropriate trade-off between safety and injury costs is through collective determinations of what acts or activities are too dangerous to be permitted. I have called this approach "collective" or "specific  deterrence." . . .

If collective deterrence were the only goal of tort law, the requirement of causal linkage would play a crucial role. Society presumably decides to forbid or otherwise restrict certain behavior because  it believes that such behavior is likely to increase the occurrence  of injuries. The safety costs, that is, the burden imposed on those who must alter their behavior in order to diminish this risk, are deemed small enough to be worth imposing. The judgment involved  in choosing to forbid or restrict particular behavior is precisely  the kind of judgment that depends on the existence of a causal  link, for there is no reason to prohibit or restrict behavior that we  do not believe will increase the chances of injury in the future. Such  behavior is, by definition, harmless to the best of our knowledge; since some people desire to engage in it, there is no reason why it  should be barred.

  Obviously, not all behavior that is causally linked to harm is worth deterring collectively. Most behavior is risky and increases  the chances of injury – that is why a trade-off between safety and  injury costs is needed. Causal linkage to injury says nothing about  the terms of the trade-off; it only defines those situations in which a trade-off may be needed. As such, the fact that an activity is  causally linked to an injury is a necessary but not a sufficient condition for the imposition of collective deterrence restrictions.  Therefore, to define the prerequisites of liability under a collective  deterrence standard it is necessary to look to other causation requirements. . . .

If specific deterrence were the only goal of tort law, collectively  proscribed behavior would be penalized regardless of whether in a specific instance it was a but for cause of harm. If drunken driving were forbidden because of its accident-causing potential, the  drunken driver would be penalized whether he had been caught as  a result of chance, of an accident, or of some other attention-getting  behavior. . . .

The requirement of proximity in causal relationships, unlike that of but for relationships, may help to select from the universe of causally linked actions those actions worth deterring collectively.  It may do so in two ways.

First, the element of foreseeability in the requirement of proximate  cause is directly relevant to collective deterrence. What, one  may ask, is the use of trying to penalize or deter collectively those acts or activities whose propensity for harm cannot be known at the time the action takes place? After the accident a causal link may  be recognized, because we now know that such acts or activities are  dangerous, but why should that be a basis for penalizing actors who  neither knew nor should have known of that risk before the accident?  Such penalties cannot alter dangerous behavior. Thus, in  terms of collective deterrence the argument for a foreseeability requirement excluding many causally linked actions from liability is  very strong. . . .

Second, the requirement of proximity may be useful in deciding which causally linked activities should be selected for control out of  the infinite number of possibilities, because proximate cause implies some look at the relative susceptibility of various actions to modifications diminishing their riskiness. It is true, as all first year torts students learn, that the requirement of proximate cause does not mean that only one among many causes must be determined to be proximate.  Yet the more subtle student also learns that whether a causally linked party is held responsible as a proximate cause  depends, in part, on what other parties were involved and how risky their behavior was in comparison to his.

If the goal of collective deterrence is to select from activities that increase the risk of injury those worth controlling (in other words, if causal linkage suggests the need for a trade-off between  injury and safety costs), then the requirement of proximity may  serve a useful function by excluding from control some relatively  less risky actions. One element in the trade-off between safety and injury costs in any particular activity is bound to be whether controlling or modifying another activity would be more efficient. The  relational aspect in proximate cause seems to look to precisely that  issue.

This discussion of proximate cause as a device for selecting relatively good avoiders of accident costs for collective deterrence  purposes may strike the reader as somewhat peculiar. After all, is  that not, at least in part, what  fault is about? Is fault not designed to select from an infinity of causally linked actions those which are by definition  not worth doing, that is, wrongful or tortious? . . . [I]ndeed, if collective deterrence were the sole goal  of tort law, it would be logical to reshape the definition of fault to  do the whole job of selecting those causally linked acts and activities collectively deemed not worth doing. . . .

To see in its full complexity the role that the concept of proximate cause plays in torts, as well as to understand, for the first time, the significance of the  requirement of a but for relationship, we must turn to another goal of tort law – market deterrence.

2. Market or General Deterrence. Although it approaches the goal differently, market deterrence aims at the same object as collective deterrence. Both seek to strike a balance between safety  and injury costs by avoiding only those injuries whose harm is sufficiently  great to justify costly avoidance. But while collective deterrence would accomplish this end through a political, collective balancing of safety and injury costs, general or market deterrence would leave the judgment to an infinity of atomistic, individual  market decisions. . . .

[A] market deterrence approach places injury costs on those  actors who can best decide whether avoidance is cheaper than bearing  those costs. The object is to place the incentives for choosing  between accident costs and their avoidance on those actors who, as a practical matter, we believe will choose most effectively. Thus, the chosen loss bearer must have better knowledge of the risks involved  and of ways of avoiding them than alternate bearers; he must be in a better position to use that knowledge efficiently to choose the cheaper alternative; and finally he must be better placed to induce modifications in the behavior of others where such modification is the cheapest way to reduce the sum of accident and safety costs.  The party who in practice best combines these not infrequently  divergent attributes is the "cheapest cost avoider" . . . .

If selection of such a cheapest cost avoider is the object of market deterrence, how helpful are the various concepts of causation  I have distinguished for accomplishing it?

Generally a causal link between an activity and an injury would  be required. . . . [H]ow can a person be the cheapest cost avoider of an injury if his actions do not increase the chances that  the injury will occur? . . . .

The role of but for causation in a system of market deterrence is less obvious than that of causal linkage. Nevertheless, market  deterrence alone among the tort goals outlined above can explain  the virtual universality of the but for test.

The function of the but for requirement in market deterrence is to assure that the injury costs allocated to the cheapest cost avoider include only those costs relevant to the choice between injury and safety. . . .  By using the but for requirement, we tell the chosen loss bearer that its burden will equal those costs that, but for its behavior, would not have been incurred; inevitably, therefore, we also tell  the loss bearer that its future insurance premiums will be based on those injury costs . . . . In this way we can approximate the optimal burden, that is, the burden that will create appropriate incentives to  avoid injuries worth avoiding and not avoid those injuries that are  too costly to eliminate. . . .

To say this, however, is not to say that the but for test should be viewed as an absolute requirement in case-by-case  determinations. Thus, where it is difficult to prove a “but for relationship"  or where either of two independent defendants was a sufficient cause of the harm and hence neither was a necessary (but for) cause, it is, at the very least, doubtful whether blind adherence to  the requirement that the victim prove a but for relationship serves the purposes of market deterrence. . . .  Functionally, the issue is this: should  the loss be allocated to the defendant in defiance of the strict but for test or is it better to let the loss fall on the wrong party, the victim, who by definition is not the cheapest cost avoider? It is little wonder that in cases of this sort courts have striven mightily, in the  face of the seemingly inexorable requirement of but for relationship, to ignore the prerequisite and place the loss correctly, if inexactly, on the defendants. . . .

General or market deterrence, then, relies on causal linkage as  one crucial element in identifying the optimal loss bearer. It relies on but for cause as a sound way of determining, on a case-by-case basis, what burden-incentive should be placed on the loss bearer.  Yet neither concept, together or separately, suffices to identify the  proper loss bearer. The requirement of proximate cause is necessary to select from actors who may be cheapest cost avoiders because they are sufficiently causally linked, those who in fact are.

Various elements in the requirement of proximate cause are  relevant to selecting the cheapest cost avoider. Foreseeability is  obviously germane, for clearly the ability to foresee risks is important  in comparing accident avoidance costs with safety costs. . . .

Past foreseeability is not, however, a prerequisite to liability if the object is market deterrence. A cheapest cost avoider of future costs exists even though no one could have foreseen what was to  transpire. Even though past injuries were not foreseeable, placing liability for them on one party rather than another may beneficially affect future choices from a market deterrence standpoint. If the  effect of such allocations is simply to categorize certain activities as  especially prone to unforeseeable and uninsurable risks, a favorable  market deterrence effect may be achieved. . . .

Proximate cause limits the degree to which even a fault system  allocates losses on the basis of fault by considering as relevant, but  not conclusive, factors that help define the cheapest cost avoiders.  In this way it serves the goal of market deterrence, which would also  consider those factors as relevant, but not decisive, and which would  not be overly impressed by the collective judgment implicit in a  finding of fault. . . .

Proximate cause, though significant in furthering market deterrence,  cannot be fully explained in terms of market deterrence,  nor can it do the whole job required by market deterrence. While   the doctrine may serve to limit the liability of faulty injurers and leave the losses on non-faulty, perhaps more proximate victims, it  cannot serve to do the reverse and burden faultless, but "very proximate"  injurers. And, if the sole object of tort law were market deterrence,  more direct ways than proximate cause could be discovered  for applying the factors relevant to selecting the best loss  bearer. But tort law does not serve a single goal. . . .

Causation, viewed as a set of functional concepts, must respond  to the sum of the goals of tort law. Causal linkage and  but for cause are functionally related to few, but fundamental tort aims. Proximate cause is, however, related to all such goals. It is little wonder  then that it has always seemed the most complex, paradoxical, but  also "flexible" and policy-based of the causal requirements. Indeed,  its very flexibility and explicitly functional policy orientation sometimes serve to hide the equally functional, rather than absolute,  roles played by the other concepts.

II.    APPLICATION OF THE MODEL TO A FEW TRADITIONAL CASES

. . . .

A.     Market Deterrence and Foreseeability of Unusual Damage

As noted above,  use of the concept of but for cause in addition to the concept of causal link can be understood only insofar as one  goal of tort law is market deterrence. Both causal elements have  traditionally been accepted as virtually absolute prerequisites to liability under the fault system, suggesting that in practice the law  has considered market deterrence to be a crucial goal of tort law. . . . [H]as the law utilized the requirement of proximate cause in a manner consistent with the promotion of  market deterrence goals? An exploration of three classic situations  dealing with one element of proximate cause – foreseeability of unusual  damage – should serve as a convenient vehicle for examining  this question.

The first case, exemplified by Palsgraf v. Long Island R.R.,involves limited or absent foreseeability of category of plaintiff. In Palsgraf railroad employees helping a passenger onto a train negligently caused him to drop a package he was carrying. Unknown to  them, the package contained fireworks. When it fell, the fireworks  exploded, causing some scales on the other side of the platform to  fall and strike Mrs. Palsgraf, injuring her. While harm to the passenger  boarding the train was a foreseeable risk of the negligent behavior, harm to Mrs. Palsgraf, who was standing well out of the way, was held not to be. At most it was only barely foreseeable that  endangering the package might harm a distant bystander.

The second case, epitomized by the "thin skull" or "eggshell plaintiff" hypothetical, involves limited or absent foreseeability of extent of damages. The defendant does something wrong. It is  deemed faulty because it is easily avoidable and likely to do the plaintiff a small, but not insignificant injury. Plaintiff has a thin skull and dies. Plaintiffs with thin skulls are foreseeable (whatever courts might say), but their presence (like the chance of harm to bystanders in Palsgraf) is sufficiently rare . . . .

The third case involves limited or absent foreseeability of what  is rather loosely termed "type" of damage. The defendant gives his three-year-old daughter a loaded gun; the daughter drops the gun on the toe of an elderly relative. Even though the gun does not go  off, the relative is seriously injured. Loaded guns are very difficult objects for three-year-olds to hold; this fact, which establishes causal linkage, is either unforeseeable or not sufficiently likely to make  handing loaded guns to infants faulty behavior. What makes the conduct faulty is, of course, the danger that the child will fire the  gun; but that did not occur.

What effect, in all these cases, should limited foreseeability or absence of foreseeability have on the proximate cause decision when that decision is viewed from a market deterrence standpoint? The  issue posed by this question would seem to be no more than: which  category, plaintiff or defendant, had the better ability to choose the safer alternative as far as this unexpected or barely expected damage  was concerned?

The fact that one party cannot foresee a certain kind of damage  does not automatically exclude it as a cheapest cost avoider, although  the chances that it is are reduced. Unforeseeable risks may,  as a statistical matter, cluster around certain activities. If those  engaging in such activities are held liable for those unforeseeable  damages, the activities will acquire a reputation for being financially risky. As such, they may be undertaken to a lesser extent or  undertaken only by those who find them especially valuable, by  people who, more than most of us, enjoy taking risks, or finally by entrepreneurs, that is, those who are willing to gamble on their  knack for avoiding even unforeseeable risks more successfully than  most people. Thus, even in the absence of foreseeability or in the  presence of only very limited foreseeability, such modifications would serve to reduce the sum of accident and safety costs, that is, to further market deterrence. . . .

Given all this, one can begin to understand how the distinctions  between barely foreseeable or unforeseeable category of plaintiff  (Palsgraf), extent of damages (the thin skull case), and type of damage (the loaded gun case) may, in a very crude sense, correspond  to distinctions made in a search for the cheapest cost avoider.

1. Foreseeability of Category of Plaintiff. The fact that we are prepared to say that the defendant could have avoided a given loss  more easily than one category of plaintiff says very little about the  relative ability of the same defendant to avoid the same loss when  compared to a totally different and perhaps unforeseeable category  of plaintiff. . . .

At the very least, the existence of an unforeseeable or only slightly foreseeable plaintiff requires the court or jury to make an  independent examination of relative cost avoidance potential between  the defendant and this plaintiff category. . . . For example, the fact  that a defendant landowner is the cheapest avoider of certain losses with respect to most categories of invitees does not mean that he is also the cheapest cost avoider with respect to many categories of  trespassers or even certain licensees.  Similarly, the fact that a construction company is the cheapest avoider of injuries to children caused by loose pipes left on a playground may, but need not, mean  that it is also the cheapest avoider of injuries to adult fools who are  also injured there.  The issue in the latter instance is whether the category "fool" is a cheaper avoider than the category "construction  company." In the given case I doubt that it would be, and I expect that a court would agree and impose liability. . . .

2. Foreseeability of Extent of Damage. . . .  By and large, if a defendant  category is the cheapest avoider of one type of damage vis-a-vis  a given category of plaintiff, then that same defendant will be  the cheapest avoider of any related, but unforeseeable or barely  foreseeable extra damages to the same category of plaintiff. . . .  Hence,  the general common law rule that, once proximate cause between  defendant's fault and the injury exists, liability extends to more  serious, but unexpected damages as well, seems to be a good starting  point from a market deterrence point of view.  It is, however, only a starting point, a rough approximation. Like the rule that liability exists only to foreseeable plaintiffs, it makes sense in terms of market deterrence only if, in practice, exceptions are made to it. . . .  

The doctrines used to limit the scope of the rule permitting  unforeseeable damages ad infinitum are not, by and large, proximate cause doctrines.  They tend instead to be introduced through  concepts of assumption of risk (in its primary sense) and, on occasion, through contributory negligence. A great violinist who goes to work in a steel mill and whose hands are mangled as a result of the mill owner's fault is not likely to recover for his "extra, unforeseeable"  damages. Many courts would say that he had assumed the risk  of such harm, because he had greater knowledge of the unusual  possible damages and could have found alternate employment less  dangerous to his hands. Yet the violinist would probably be able to  recover for the same excess damage to his hands if they were injured  in a car accident. He might have greater knowledge of the particular  risk involved than the defendant could have, but that knowledge  would not provide him with any meaningful alternatives.  The defendant  in this situation, unlike the defendant in the steel mill case, would, if he were the cheapest avoider of the original injury, also be  the cheapest avoider of the unlikely extra damages.

The same types of distinctions that limit and refine the rule  permitting unforeseeable damages so that it serves market deterrence goals can be seen in other classic hypotheticals as well. If the  package belonging to the passenger in Palsgraf had contained a Ming vase, the defendant probably would not have been held liable for the extraordinary damages caused by the breakage, despite the basic common law rule imposing liability and despite the fact that the risk that made the defendant's behavior faulty in the first place  was precisely the risk of damage to the package.  Some courts would have found contributory negligence in this situation, others  assumption of risk; whatever the terminology, the result appears  foreordained and in accord with market deterrence goals. Conversely, the person with a thin skull generally recovers for all his injuries even though his knowledge of the condition must be vastly  superior to the defendant's. Knowledge without alternatives (adequate  helmets do not exist; if they did, the cases might well be  different) does not make a category of actor a cheap cost avoider. . . .

3. Foreseeability of Type of Damage. Ability to foresee category of plaintiff, as the common law defines it, is highly relevant to selecting cheapest cost avoiders. Ability to foresee extent of damages is much less relevant . . . Ability to foresee type of damages . . . lies somewhere in between.

The example used above for unforeseeable or only slightly foreseeable  type of injuries was the case in which an adult handed a loaded gun to an infant. The child drops it on the toe of an elderly  person who is seriously injured even though the gun does not go off.  It is known, either before or after the accident, that loaded guns are  particularly hard for infants to hold. Causal linkage is, therefore, established because the evidence indicates that handing loaded  guns to children substantially increases the risk that they will be dropped and thus the risk that elderly feet will be injured as a result.

Our initial, instinctive reaction to this rather outlandish hypothetical  is to want to hold the negligent defendant liable. That  reaction, I would suggest, is based on collective deterrence grounds.  The foreseeable harm from handing an infant a loaded gun is enormous;  the harm that occurred is certainly no greater than the expected  harm; the collective deterrence penalty appropriate to the  defendant would be at least as great as the damages suffered; then  why not hold the defendant liable?. . . .

This collective deterrence effect may be demonstrated by contrasting  a variation on Palsgraf with the hypothetical case. In  Palsgraf the defendant's negligence consisted of unnecessarily endangering the package. Suppose that the owner of the package had  had a "thin shoulder" and that, while shoving him onto the train in a non-negligent manner except with respect to the package, the  railroad employees jostled his arm so that it was severely injured  and had to be amputated. Here, unlike the gun hypothetical, there  is no collective deterrence equivalence between the wrong committed  and the penalty to be imposed; hence, our instinct is to say "no  liability." . . . .

The circumstances surrounding the occurrence of an unexpected  type of damage may suggest that the defendant who is the best bearer of the basic loss for market deterrence purposes is not  the best bearer of the different type of loss. But they may suggest  the opposite. Or they may simply leave us perplexed on the point  and hence more prone to follow the dictates of other goals (whether  collective deterrence as in the "loaded gun" case or compensation),  which may seem to be clearly furthered by one result or another. It  is not surprising, therefore, that such cases appear at first glance to  be the most mixed up of all the "unexpected damage" cases. . . .

III.    THE LANGUAGE OF CAUSATION AND THE GOALS OF TORT LAW

Causal requirements, like all other legal requirements, must ultimately justify themselves in functional terms. Law is a human  construct designed to accomplish certain goals. . . .

In this sense many seemingly significant philosophical questions  concerning cause become irrelevant to the use of that term in law. To amplify: so far as legal language is concerned, the "cause"  of a disease would depend on how, at any given time, it could be  most easily controlled. From this point of view, in the nineteenth  century it would have been appropriate to speak of the "cause" of tuberculosis as the absence of sun and the presence of bad living conditions. Other possible factors, that were both but for causes and causally linked, were not subject to human control, current or even hypothetical. With the identification of the Koch bacillus all that  changed. At first potentially, and subsequently in practice, efforts directed at this causally linked element seemed most likely to control the disease. It was unimportant that both living conditions and  genetic predisposition also mattered. What use would it be to speak  of these elements as causes when living conditions were so hard to modify in comparison to controlling the bacillus and when genetic  predisposition was totally outside human control? Yet these two factors remained causally linked to tuberculosis. More recently, the  prospect of genetic engineering has again changed the causal language  appropriate to this disease. Now one can, in a meaningful  way, speak of genetic. predisposition as a "cause" of tuberculosis.

The objective, with respect to tuberculosis, is easier to define  than are most legal objectives or goals. Causal language in law must  respond to the fact that at times the object sought may differ while  the factors subject to control remain the same. For example, courts  have been able to hold two faulty, independent hunters both liable  in cases like Summers v. Tice, even though only one of their shots hit the victim and even though a showing of but for for cause was supposedly a prerequisite for liability. They would not as readily have found two equally faulty independent rapists to be the fathers  of an illegitimate child – even though the causal evidence as to the  wrongdoing was precisely the same as in Summers v. Tice and even though the victim was equally innocent. Rightly or wrongly, the  effects (and hence the function) of paternity actions are conceived to be very different from those of money damage claims for injuries.  What is "cause" for one need not be "cause" for the other.

If we recognize that in law the term "cause" is used in different guises but always to identify those pressure points that are most  amenable to the social goals we wish to accomplish, the apparent paradox disappears. Where goals differ, so does the practical definition  of causation. Some aspects of causation (like proximate cause)  show this functional basis on the surface. Others, like causal linkage  and  but for relationship, often seem to be almost absolute, goal-neutral requirements. But they appear that way simply because  they are being employed in contexts where, by and large, the appropriate goals require them to be established almost universally. However, the willingness of courts to ignore not only but for causes in cases like Summers v. Tice,  but even causal linkage where extreme spreading differences seem involved indicates, I suggest, that in the  law "cause in fact" (as it was once called), like proximate cause, is  in the end a functional concept designed to achieve human goals.

If causal concepts can be used flexibly to identify the pressure  points most amenable to our social goals, then use of such concepts  has great advantages over explicit identification and separation of  the goals. Terms with an historical, common law gloss permit us to  consider goals (like spreading) that we do not want to spell out or  too obviously assign to judicial institutions. Because, like all moral  terms, causal terms have come to have meanings of their own that  cannot be changed as a result of one person's analysis, they enable  us to resist political pressures that, would, if a more "goal conscious,"  antiseptic language were employed, result in a mixture of  goals thought to be less desirable. Finally, and probably most importantly,  they enable the introduction of goals we have not been  able to spell out or to analyze, but which nonetheless, together with  analyzed goals, form part of that set of relationships we call "justice." . . .

As in the past, causal  concepts seem to be responding to changes in appropriate pressure  points, yet doing so without breaking with the past or with any not  clearly understood goals the past contains. . . .